Pulmonary nodule malignancy probability: a meta-analysis of the Brock model

This study aims to quantify the performance of the Brock model through a systematic review and meta-analysis and to clarify its overall accuracy in predicting malignant pulmonary nodules.

A systematic search was conducted in databases including the Cochrane Library, Excerpta Medica database (EMBASE), MEDLINE, Web of Science, Chinese Biological Medicine Database (CBM), China National Knowledge Infrastructure (CNKI), VIP, and Wanfang from their inception until May 1, 2024, to collect observational cohort studies involving the Brock model. The primary outcome was the pooled area under the receiver operating characteristic curve (ROC) the area under curve (AUC) for the Brock model. Secondary outcomes included sensitivity and specificity. The metaprotocol was registered in the International Prospective Register of Systematic Reviews (CRD42024538163).

A total of 52 studies involving 85,558 patients were included. The pooled AUC was 0.796 (95% confidence interval [CI]: 0.771-0.820), with a pooled sensitivity of 0.82 (95% CI: 0.76-0.87) and specificity of 0.80 (95% CI: 0.72-0.86). Subgroup analysis showed that the performance of the full model was significantly better than that of the simplified model (0.822, 95% CI: 0.794-0.849 versus 0.687, 95% CI: 0.611-0.763). The model performed excellently for pulmonary nodules with diameters of 1- to 8 mm (AUC: 0.927, 95% CI: 0.900-0.954). However, its performance was lower in Asian populations (AUC = 0.741, 95% CI: 0.703-0.780), solitary pulmonary nodules (AUC = 0.767, 95% CI: 0.693-0.842), and subsolid pulmonary nodules (AUC = 0.747, 95% CI: 0.661-0.832).

This meta-analysis confirms the Brock model's overall strong performance. However, the results indicate certain application limitations of the Brock model, with reduced accuracy for larger nodules (>15 mm), solitary pulmonary nodules, subsolid nodules, and in Asian populations.